The Relationship Between Teachers' Self-Efficacy Beliefs and Job Satisfaction Levels: A Meta-Analysis Study

In this study, it is aimed to examine the relationship between teachers' self-efficacy beliefs and job satisfaction levels with meta-analysis method. For this purpose, it has been investigated whether there is a significant difference between the effect sizes of the studies examining the relationships between teachers’ self-efficacy beliefs and job satisfaction levels according to the moderator variables. The studies examining the relationship between teachers' self-efficacy beliefs and job satisfaction levels in the years 2000- 2018 including the number of samples and correlation coefficient (or t value) were included in the study. The studies within the scope of this study was accessed by scanning the YOK National Thesis Center, Google Academic and Proquest databases. As a result of the literature review, 35 studies which were in accordance with the selection criteria of the research were included in the meta-analysis. In the study, “ Fisher’s z”  value was used to calculate the effect size values. The effect direction and overall effect size of the studies within the scope of the study were calculated according to the random effects model. As a result of the analysis, it was determined that the overall effect size of the relationship between teachers' self-efficacy beliefs and job satisfaction levels was positive and moderate. In addition, it was observed that there was a significant difference between the effect sizes of the studies examining the relationships between teachers' self-efficacy beliefs and job satisfaction levels acoording to the moderator variables (study year, study location, type of school and type of scales used in the studies).


Introduction
Job has a basic function in achieving the economic gain required for the survival of individuals (Eğinli, 2009). Considering that a working individual spends a large part of his time in the workplace for 25-30 years, his job satisfaction is as important as his job in terms of his physical and psychological health (Telman & Ünsal, 2004). Job satisfaction is defined as the attitude or the pleasure, satisfaction or happiness he / she has developed as a result of the perceptions of the individual about his / her work (Davis, 1988;Gedik & Üstüner, 2017;Keser, 2005). According to another definition, job satisfaction is an attitude, pleasure or emotional situation that the employee reaches as a result of evaluating himself / herself and his / her job (Gümüş, 2017). Job satisfaction is seen as one of the most important requirements of individuals to be happy, successful and productive (Günbayı & Tokel, 2012). A high level of job satisfaction poses positive effects on the physical and mental health of the employees while low job satisfaction can cause various physical illnesses (shortness of breath, headache, loss of appetite, nausea, etc.) and psychological problems (anxiety, burnout, etc.) (Türk, 2007). It is known that job satisfaction affects not only the individual but also the organization. High job satisfaction leads to desirable results by the organization, such as the rate of labor force turnover and absenteeism, organizational commitment, and high productivity due to increased organizational citizenship behavior (Akomolafe & Olatomide, 2013;Çetinkanat, 2000;Ghazzawi, 2008;Ghazzawi & Smith, 2009;Kristof, 1996;Oyewobi, Suleiman, & Muhammad-Jamil, 2012;Robbins & Judge, 2012;Verquer, Beehr, & Wagner, 2003). But, the low level of job satisfaction may lead to negative outcomes such as quitting, absenteeism or poor performance (Feldman & Arnold, 1983).
As in all organizations, job satisfaction is of great importance in education organizations. The condition that teachers can successfully fulfill the duties and responsibilities imposed on them is related to the satisfaction they receive from their profession (Gençtürk & Memiş, 2010). Job satisfaction for teachers can be defined as the attitude of the teacher towards his / her students and his / her school "or "the satisfaction or dissatisfaction of teachers from their work" (Vural, 2004, p.32). The high level of job satisfaction of the teachers affects the realization of the educational objectives in a positive way. A school with high job satisfaction teachers is expected to provide quality education and to train successful students (Demirtaş, 2010). Indeed, findings of studies examining the relationship between teacher job satisfaction and student achievement confirm this expectation (Crawford, 2017;Secumski Kiligian, 1993;Westfall, 2012). On the other hand, it is possible for teachers with low job satisfaction to develop negative attitudes towards themselves and their students, and students to develop negative attitudes towards the lesson, and also to go to work as little as possible and even think about leaving the profession (Büyükşahin Çevik, 2017;Cunningham, 2015;Jackson, 2018;McCaffery, 1976).
The individual and organizational results of job satisfaction play an important role in achieving the goals of the organization. Therefore, knowledge of factors affecting job satisfaction may reveal what should be done to increase job satisfaction. The rate of meeting these expectations, which can be varied in many areas such as wage, promotion, social security, job security, quality of business relations, are the organizational factors that constitute job satisfaction (Aşık, 2010). However, it is known that job satisfaction is related to these organizational factors as well as individual factors (age, gender, education level, locus of control, optimism, self-efficacy belief, etc.) (Aşık, 2010;Rauf, 2010;Rothmann, 2001;Shoji et al., 2016). Some organizational (nature of the work, structure of the organization) and individual factors (age, gender) can be difficult or impossible to change. Self-efficacy of individual factors is a factor that can be changed and has a significant effect on job satisfaction. Self-efficacy is the self-belief of a person about his capacity to perform successfully, by organizing the activities necessary to show a certain performance (Bandura, 1997). Self-efficacy is also expressed as a person's confidence in his / her own performance (Açıkgöz, 1996). Self-efficacy belief affects the individual's thoughts, goals, lifestyle, efforts in the face of difficulties and products that will be the result of efforts (Bandura, 2001). If the individual who is facing any difficulty has serious concerns about his or her abilities, this individual may slow down or cease to do so in order to eliminate the hardship. On the other hand, the individual who has full confidence in his / her abilities will be more determined to solve the problem in the event of difficulty, and will be determined to resolve it (Hazır Bıkmaz, 2002). In this respect, the self-efficacy of the individual in relation to his work will affect the actions, organizational performance and efficiency of overcoming the problems to be encountered in the workplace.
Self-efficacy has attracted the attention of education researchers recently. Teachers' self-efficacy is defined as the belief that teachers have in their ability to have a positive impact on student learning (Ashton, 1984). Teachers who are satisfied with their jobs usually have a high degree of professional competence. Self-efficacy beliefs and job satisfaction levels of the teachers who feel themselves qualified in terms of their knowledge of subject matter, teaching skills and secured about classroom management are high (Akomolafe & Ogunmakin, 2014;Wang, 2013). Teachers' personal judgment on their abilities and skills is shaping the quality of education. Teachers with a strong sense of self-efficacy are better planned to teach instruction, tend to show higher performance in the teaching process and are more open to new ideas, and are more willing to try new methods to better meet the needs of their students (Saraçaloğlu, Aldan, Karademir, Dinçer, & Dedebali, 2017). Teachers with high self-efficacy believe that students can influence their success (Armor et al., 1988, as cited in Chesnut & Burley, 2015. These teachers devote more time to academic learning in the classroom and learning of students with learning difficulties and motivates the student by praising even the smallest achievements of their students (Gibson & Dembo, 1984). Teachers with high self-efficacy are more likely to have students with higher achievement scores (Allinder, 1995;Ashton, 1984;Brown, 2012;Eberle, 2011;Ross, 1992;Sheftall, 2000). On the other hand, teachers with low self-efficacy have a negative effect on stundents' self-efficacy beliefs and cognitive development by blaming the them for their failures (Gibson & Dembo, 1984).
There are many studies that demonstrate the relationships between self-efficacy beliefs and job satisfaction levels (Blackburn, 2008;Buluç & Demir, 2015;Canrinus, Helms Lorenz, Beijaard, Buitink, & Hofman, 2012;Caprara, Barbaranelli, Steca, & Malone, 2006;Federici & Skaalvik, 2012;Türkoğlu, Cansoy, & Parlar, 2017). In fact, it is possible to find a meta-analysis of these relations in the literature (Judge & Bono, 2001). Judge and Bono (2001) conducted a meta-analysis of the relationships between basic self-evaluation traits (self-efficacy, self-esteem, emotional stability), job satisfaction and job performance. The study presents a limited systematic analysis of the relationship between self-efficacy beliefs and job satisfaction levels. The sample of the studies included in the meta-analysis is not composed of teachers but different organizations. The study does not provide sufficient information about the current situation since some of the meta-analysis studies belong to the old times. In addition, the study did not reveal the effect of potential moderator variables (type of publication, study year, study location, type of school and type of scales used in the studies) on self-efficacy and job satisfaction relationship. However, moderator variables may have some effect on this relationship: a) Graduate thesis studies are carried out on large samples with different characteristics compared to the papers in general. This situation may cause differences in terms of research results.
b) The concepts of self-efficacy and job satisfaction were first put forward in the United States and the studies examining the relations between these two concepts have been carried out widely in different countries. The meanings attributed to the concepts may vary culturally and nationally. This may be reflected in the results of the scale preparation processes indirectly.
c) Teachers begin their jobs with certain professional qualifications. Over time, these qualifications may be insufficient to raise the required human profile. In accordance with the conditions of the day, the teacher may update his / her professional competencies, develop or acquire new qualifications. The teacher's ability to provide this professional development or not may have an effect on self-efficacy beliefs and thus job satisfaction levels.
d) In the studies conducted to determine whether teachers' self-efficacy beliefs and job satisfaction levels differ with various variables, the type of school appears to be an important variable (Aslan & Kalkan, 2018;Bümen, 2009;Karakaya-Çiçek & Çoruk, 2017;Koruklu, Feyzioğlu, Özenoğlu Kiremit, & Aladağ, 2013;Liu, 2008). However, the effect of the school type on the relationship is unknown. Therefore, the effect of the school type on the relationship needs to be determined.
These limitations and rationales reveal the need to conduct an up-to-date and comprehensive meta-analysis study by synthesizing studies examining the relationships between teachers' self-efficacy beliefs and job satisfaction levels. A meta-analysis study on the subject is important in terms of revealing the direction and level of the relationship between self-efficacy and job satisfaction. Investigation of the effect of moderator variables on the relationship will provide important information on the field literature on whether relationship varies by depending on type of publication, study year, study location (in or outside Turkey), type of school where the teachers work and type of scales used in the studies or if change is provided, how this change occurs. This information can help researchers to assess what kind of studies can be carried out at individual and organizational levels to strengthen the relationship between teachers' self-efficacy beliefs and job satisfaction levels, or which moderators should be considered in future studies. The aim of this study is to investigate the relationship between teachers' self-efficacy beliefs and job satisfaction levels with meta-analysis method. For this purpose, the following sub-problems were sought: 1. What are the frequency and percentage distribution of the moderator variables of the studies included in the meta-analysis? 2. What is the overall effect size of the relationship between teachers' self-efficacy beliefs and job satisfaction levels? 3. In the literature, is there a significant difference between the effect sizes of studies examining teachers' self-efficacy beliefs and job satisfaction levels according to moderator variables (type of publication, study year, study location, type of school and type of scales used in the study)?

Method
Meta-analysis process was formed by taking into consideration the operations to be done before the analysis of the data and the issues to be considered in the analysis of the data. (i) selection of data (studies), (ii) preparing the data for the analysis, (iii) determining the analysis model, (iv) interpreting the results of the analysis.

i) Selection of Data (Studies)
In meta-analysis, it is important to expose the criteria used in determining the studies to be included in the research in the meta-analysis protocol and to be consistent with the purpose of the research in order to prevent publication bias (Berman & Parker, 2002). In this research, the following criteria were taken into account in determining the studies to be included in the scope of meta-analysis:

Criterion 1: Database of studies
In line with the aim of the study, all studies that give the relationship between teachers' selfefficacy and job satisfaction perceptions were scanned in YÖK National Thesis Center, Proquest and Google Scholar databases. The studies carried out in Turkey were searched with the keywords of "öz yeterlik" and "iş doyumu" in the YÖK National Thesis Center and Google Scholar databases. The studies conducted outside Turkey were searched with keywords of "self efficacy" and "job satisfaction" in Proquest and Google Scholar databases. As a result of the screening, a total of 61 studies, 17 of which were conducted in Turkey and 44 of which were conducted outside of Turkey, revealed the relationship between self-efficacy and job satisfaction.

Criterion 2: Confomity of the studies in terms of method
In the meta-analysis studies, it was taken into consideration that empirical studies conducted in the years 2000-2018, and taken over the teachers who worked in preschool and compulsory education institutions (primary, secondary or high school) were conducted in order to reach the effect size. As a result of examining the samples and methods of the studies, total of 55 studies, 15 of which were conducted in Turkey and 40 of which were conducted outside of Turkey, were found to meet the required criteria.
Criterion 3: Availability of statistical data of studies In order to calculate the required effect sizes in the meta-analysis study, sample size should be given with correlation coefficient between dependent and independent variables. Studies including sample size and t values obtained from regression analysis and which allow calculation of correlation coefficient were also included in the study. However, only studies that provide correlation coefficients or t-values of the sub-dimensions of the variables and whose general correlation coefficient cannot be calculated are not included in the study. As a result of examination, a total of 34 studies, 13 of which were conducted in Turkey and 21 of which were conducted outside of Turkey, meet the required criteria revealed the relationship between self-efficacy and job satisfaction. However, in the study of , the sample was divided into two groups as private and public teachers and different results were obtained for both sample groups. In this study, the study of  was evaluated as two different studies. Therefore, the number of studies with numerical data required for meta-analysis was determined to be 35.

ii) Preparing Data for the Analysis
In order to conduct a meta-analysis study, individual study weights should have a balanced effect on the overall effect size of the study, studies should not cause publication bias and effect sizes should exhibit normal distribution. Therefore, prior to the meta-analysis of the studies, the creation of the coding form should be prepared, the individual effect size of each study should be calculated and studies that have a significant impact on the overall effect size, cause publication bias and adversely affect the normal distribution of data should be identified, examined and decided to be excluded from the scope of meta-analysis.

Creation of the coding form
The coding form developed by the researcher was used in the coding of the data obtained after the selection of the data. The coding form consists of three parts; study ID, study content and study data. The identity of the study includes information about the author or authors, year and type of the study. The study content section contains information on where the study was conducted, the type of school and the type of scales used in the study. In the study data section, the numerical data required to calculate the effect size; sample size and correlation coefficient (or t value) are given. In accordance with the information obtained from the coding form, the moderator variables of the study were determined as follows (an example of the coding form of the studies included in the meta-analysis is given in Appendix 1

Calculation of effect size of studies
The effect size (Cohen, 1988) that can be calculated based on arithmetic averages, ratios and correlations can be used if the difference between two variables or two groups is quantitative (Borenstein, Hedges, Higgins, & Rohtstein, 2013). Correlations are used to calculate effect sizes for relational survey models. In this study based on a meta-analysis of relational survey models, the effect size values and the combined effect size and publication bias of each study included in the meta-analysis were calculated using the Comprehensive Meta-analysis v2.2.064 (CMA) Statistical Package program. SPSS package program was used in the graphs of the normal distribution of the data.

Examination of the distribution of the studies
In order to combine the effect sizes obtained from the studies in the meta-analysis, it is necessary to determine whether they have a normal distribution for these effect sizes. Normal Q-Q graph, Shapiro-Wilk test results, skewnes and kurtosis coefficients were used to determine whether the effect sizes of the 35 studies considered to be included in the meta-analysis have normal distribution. Normal Q-Q graph shows the relationship between the theoretical (expected) normal distribution values and the actual values observed. When the values of the theoretical distribution coincide with the actual values, it is a right that makes 45 degrees open. However, the decision as to whether the Q-Q graph is linear is subjective (Can, 2013). The distribution of the effect sizes of 35 studies is given in Graph 1.

Graph 1. Normal Distribution Graph of Effect Sizes
When the graph 1 is examined, it is seen that the effect sizes of 35 studies included in the study are collected along a straight line. However, in order to be able to decide on the normality of the distribution, the Shapiro-Wilk test results and the skewness and kurtosis coefficients were examined. According to the Shapiro-Wilk test results, the effect size values show normal distribution (S-W = 138; p > 0,05). In addition, the skewness (0,46) and kurtosis (1,01) values of the effect sizes are in the normal distribution range (+1,96 and -1,96). In line with this statistical information, it is considered appropriate to combine the effect sizes of 35 studies for meta-analysis.

Calculation of study weights
When the numerical data of the studies are examined in detail, some of the studies on the larger sample group (Aldridge & Fraser, 2016, n = 789; Caprara et al., 2006Caprara et al., , n = 2184Skaalvik & Skaalvik, 2014, n = 2569 or the smaller group (Güngör Seyhan, 2015, n = 52; Rincon, 2018, n = 14; Ruma, Houchins, Jolivette, & Benson, 2010, n = 70) were obtained. In the meta-analysis it is desirable that the studies included in the analysis are similar. In a meta-analysis studies with small sample size, if the effect sizes of the study with a large sample size is significantly different from other studies or if it significantly differentiate the overall effect size, it is useful to exclude the study with a large sample size from the analysis and and explain in detail (Dinçer, 2014). In order not to cause false interpretations, the weight of these studies according to the random effects model was examined. In addition, changes in the overall impact coefficient were investigated when the studies were included in the research. It was decided that these studies should not be excluded from the study as the weights of the studies were close to each other and did not change the effect size significantly if they were included in the research.

Examination of publication bias of studies
When the literature is examined, the studies with statistically significant results are more likely to be published. As the statistically non-significant results tend to have a small effect size, studies with large effect size are more likely to be published than those with a small effect size (Bakioğlu & Göktaş 2018;Üstün & Eryılmaz, 2014). Since it is more preferable to include published studies in meta-analysis, the possible biases in these studies are also reflected in the meta-analysis. This problem is generally referred to as "publication bias" (Bakioğlu & Göktaş 2018). Therefore, it should be tested whether the studies cause publication bias. In this research, publication bias was tested by funnel scatter plot, Rosenthal's Safe N Test, Begg and Mazumdar rank correlations were tested by Egger's Linear Regression method.
In the interpretation of the funnel scatter plot, the location of the individual effect sizes of the studies included in the meta-analysis is an important factor. The effect sizes of the studies are all in the funnel lines and they are observed in symmetrical form when there is no publication bias. The middle line shows the overall effect and studies are expected to focus on the overall effect of the individual effect sizes. Studies whose effect sizes are not included in the funnel may cause publication bias (Dinçer, 2014). The scattering of the effect sizes of the studies included in the scope of meta analysis is given in Chart 2.

Graph. 2. Funnel Scatter Plot
As can be seen in Graph 2, the majority of the effect sizes of the studies included in the metaanalysis were collected in a funnel and symmetrically. In addition, it is seen from the Graph 2 that the individual effect sizes of the studies are gathered around the middle line showing overall effect size. According to the funnel scattering graph, 35 studies included in the meta-analysis do not have publication bias. However, publication bias statistics should also be considered since the individual effect sizes of the studies are not symmetrical in the funnel. Confidence tests and results showing the bias of the studies included in the meta-analysis are given in Table 1. As seen in Table 1, Rosenthal's Safe N Test results reveal that the meta-analysis result is statistically significant (p = 0,000). In other words, in order to eliminate the significance of meta-analysis results, 2147 studies with zero effect size value are needed. The fact that Kendall's Tau coefficient obtained from Begg and Mazumdar Rank Correlations is not statistically significant (-0,035 and p = 0,776) is an indication that there is no publication bias. It can be said with 95% confidence that there is no publication bias from the result of Egger's Linear Regression method (p = 0,769> 0.05).

iii) Determining the analysis model
In determining the model in meta-analysis studies, it can be tested whether the effect sizes are homogeneously distributed. According to the results of this test, if the effect sizes are homogeneous distribution, it is said that fixed effects model can be used. If effect sizes do not show homogeneous distribution, random effect model should be used (Ellis, 2010, as cited in Gözüyeşil & Dikici, 2014). In meta-analysis studies, the choice of model should be decided as "a priori" before the analysis (Başol, 2016). The studies in the current research based on social sciences, in different countries and educational levels, and the variety of patterns and scale used in the analysis shows that the random effects model for meta-analysis is more appropriate. The Homogeneity Test was applied to the data obtained from the studies to determine the meta-analysis model based on the distribution of effect sizes.
Q value obtained as a result of homogeneity test is statistically significant [Q = 656,412 (p = 0,000)]. In the χ2 table, the 34 degrees of freedom value was found to be 48,602 at 95% significance level. The Q statistic value (Q = 656,412) was greater than the critical value of χ 2 distribution with 34 degrees of freedom [χ 2 (0,95) = 48,602] and the p value (p = 0,000) being less than 0,05 was used to determine the effect size distribution. The distribution of effect sizes indicates that it is heterogeneous. However, in the meta-analysis studies conducted with a small number of studies, Q statistic is weak in accurately determining the heterogeneity between studies included in the meta-analysis (Huedo Medina, Sánchez Meca, Marín Martínez, & Botella, 2006). Therefore, I² value was also examined to determine whether the heterogeneity between studies actually existed. I², which is not sensitive to the number of studies and the effect size, is a useful measure in terms of showing the true homogeneity ratio of the total change in the observed effect (Borenstein et al., 2013). As a result of the homogeneity test, I² value was calculated as 94,820%. Based on this value, the percentage of total variability that can be attributed to true heterogeneity or variability between studies can be expressed as 94.820%. Higgins and Thompson's (2002) I² values show 25% (I² = 25) low, 50% (I² = 50) moderate and 75% (I² = 75) high heterogeneity. According to this classification, it can be said that the calculated I² value represents 94,820% (I² = 94,820) high heterogeneity. In addition, p value (p = 0,000) is less than the significance value of 0,05 (p > 0,05). All these values (Q = 656,412, p <0.05, I² = 94,820) indicate that the distribution of effect sizes is heterogeneous and the use of random effects model is more suitable for interpreting effect sizes.

iv) Interpreting the results of the analysis
In the meta-analysis studies, the correlation values are converted to Fisher z value and the analyzes are carried out on these values. In the evaluation of the analysis findings, it is interpreted by converting to correlation coefficient (Önder & Tulunay Ateş, 2017). In the correlation data, the correlation coefficient is used as the effect size in relation to the direction of the relationship (positive or negative). If their effect sizes were between ± 0,00 and ± 0,10, ± 0,10 and 0,30, ±0,30 and ± 0,50, ± 0,50 and ± 0,80, ± 0,80 and above, the correlation coefficients were interpreted as very weak effect, weak effect, medium effect, strong effect, very strong, respectively (Cohen, Manion, & Morrison, 2007).
The question of "What are the overall effect size of the relationship between teachers' selfefficacy beliefs and job satisfaction levels?" is determined as the second sub-problem of the study. Accordingly, the weight, the effect size of each study included in the meta-analysis and the overall effect size are given in Table 3.  As can be seen from Table 3, the overall effect size value for the relationship between selfefficacy and job satisfaction is determined as 0,359 based on the random model. According to the classification of Cohen et al. (2007) is a medium effect. According to random effects model, the lower limit of the effect size is 0,278 and the upper limit is 0,440 in 95% confidence interval. The values of effect sizes were statistically significant (z = 8,665; p = 0,000). According to these findings, it can be said that there is a moderate and positive relationship between teachers' self-efficacy and job satisfaction relationship mean.
The question of "In the literature, is there a significant difference between the effect sizes of studies examining teachers' self-efficacy beliefs and job satisfaction levels according to moderator variables (type of publication, study year, study location, type of school and type of scales used in the study)" is determined as the third sub-problem of the study. In this respect, the Q-test-homogeneity test was conducted according to the moderator variables and the results were presented as tables. The effect sizes distribution and homogeneity test results are given in Table 4 related to the publication type of the studies. In order to determine the effect of publication type on the overall effect size, the studies was divided into two different groups, including thesis and paper. Related to the type of publication; the effect size value of the thesis type (0,411) was found to be greater than the effect size value (0,377) of the paper type. Homogeneity test value was found as QB = 1,312. The QB statistic value [QB = 1,312; p = 0,252] is not significant as it is below the critical value [χ 2 (0,95) = 3,841] of the χ 2 distribution at 95% significance level with one degree of freedom. Accordingly, the homogeneity hypothesis of the distribution of effect sizes was accepted within the framework of the fixed effects model. It can be stated that there is no significant difference between the effect sizes of the studies included in the scope of meta analysis related to the type of publication. Table 5 shows the distribution of the effect size and homogeneity test results related to the study year. In order to determine the effect of study year on the overall effect size, studies was divided into eight different groups, including 2010, 2011, 2013, 2014, 2015, 2016, 2017, 2018. As at least two studies are required for each group created in the calculation of the effect size, 2000, 2006 and 2012 with less than 2 two studies has been removed. Related to study year, the greatest effect size value (0,821) was seen for 2016; the lowest effect size value (0,299) was calculated for 2015. Homogenity test was found as QB =172,452. The QB statistic value [QB = 172,452; p = 0,000] is significant as it is above the critical value [χ 2 (0,95) = 14,067] of the χ 2 distribution at 95% significance level with seven degree of freedom. Accordingly, the homogeneity hypothesis of the distribution of effect sizes was accepted within the framework of the random effects model. It can be stated that there is a significant difference between the effect sizes related to the years in which the studies included in the scope of meta-analysis were conducted. Table 6 presents the results of the distribution of the effect size and homogeneity of the test related to the study location. In order to determine the effect of study location on the total effect size, the studies were divided into two groups as in Turkey and outside Turkey. Related to the study location, the effect size value of the studies conducted outside Turkey (0,400) was found to be greater than the effect size value of the studies conducted inTurkey (0,331). Homogeneity test value was found as QB = 12,529. The QB statistic value [QB = 12,529; p = 0,000] is significant as it is above the critical value [χ 2 (0,95) = 3,841] of the χ 2 distribution at 95% significance level with one degree of freedom. Accordingly, the homogeneity hypothesis of the distribution of effect sizes was determined as the random effects model. It can be stated that there is a significant difference between the effect sizes related to the place where the studies included in the meta-analysis are carried out. Table 7 shows the distribution of the effect size and homogeneity test results related to type of school. In order to determine the effect of type of school on the overall effect size, studies was divided into eight different groups, including Pre. S., P. S., S. S., H. S., P. S. + S. S., S. S. + H. S., P. S. + S. S. + H. S. and Pre. S. + P. S. + S. S. + H. S. As at least two studies are required for each group created in the calculation of the effect size, Pre. S. + P. S. and Pre. S. + P. S. + S. S. + H. S. + Unstated with less than 2 two studies has been removed. Related to type of school, the greatest effect size value (0,436) was observed for P. S. ; the lowest effect size value (0,169) was calculated for S. S. Homogenity test was calculated as QB = 80,963. The QB statistic value [QB =80,963; p = 0,000] is significant as it is above the critical value [χ 2 (0,95) = 14,067] of the χ 2 distribution at 95% significance level with seven degree of freedom. Accordingly, the homogeneity hypothesis of the distribution of effect sizes was accepted within the framework of the random effects model. It can be stated that there is a significant difference between the effect sizes related to the type of school in which the studies included in the meta-analysis are carried out. Table 8 presents the results of the distribution of the effect size and homogeneity of the test relate to type of self efficy scale. In order to determine the effect of type of self efficacy on the overall effect size, studies was divided four different groups, including Caprara et al. (2006), Gibson and Dembo (1984), Sherer et al. (1982), Tschannen Moran and Woolfolk Hoy (2001). As at least two studies are required for each group created in the calculation of the effect size, developed scales with less than 2 two studies has been removed. According to the type of self-effcacy scale used in the studies; the greatest effect size value (0,377) was observed for the scale devepoled by Tschannen Moran and Woolfolk Hoy (2001); the lowest effect size value (0,206) was calculated for the scale devepoled by Sherer et al. (1982). Homogenity test was calculated as QB = 24,531. The QB statistic value [QB = 24,531, p = 0,000] is significant because the χ 2 distribution is above the critical value [χ 2 (0,95) = 7,815] with three degrees of freedom and 95% significance level. Accordingly, the homogeneity hypothesis of the distribution of effect sizes was accepted within the framework of the random effects model. It can be stated that there is a significant difference between the effect sizes of the studies included in the meta-analysis in related to the type of self-efficacy scale used in data collection. Table 9 shows the distribution of the effect size and homogeneity test results related to the type of job satisfaction scale.  Weiss et al. (1967) 7 0,333 0,287 0,379 Smith et al. (1969) 2 0,291 0,252 0,331 Spector (1985) 4 0,103 0,015 0,191 Warr et al. (1979) 2 0,518 0,413 0,623 In order to determine the effect of type of job satisfaction on the overall effect size, studies was divided five different groups, including Brayfield and Rothe (1951), Weiss et al. (1967), Smith et al. (1969), Spector (1985) and Warr et al. (1979). As at least two studies are required for each group created in the calculation of the effect size, developed scales with less than 2 two studies has been removed. Related to the type of job satisfaction scale used in the studies, the greatest effect size value (0,518) was calculated for the scale devepoled by Warr et al. (1979); the lowest effect size value (0,103) was observed for the scale devepoled by Spector (1985). Homogenity test was calculated as QB =39,016. The QB statistic value [QB = 39,016, p = 0,000] is significant because the χ 2 distribution is above the critical value [χ 2 (0,95) = 9,488] with four degrees of freedom and 95% significance level. Accordingly, the homogeneity hypothesis of the distribution of effect sizes was accepted within the framework of the random effects model. It can be stated that there is a significant difference between the effect sizes of the studies included in the meta-analysis related to the job satisfaction scale used in data collection.

Discussion and Conclusion
In this study, which was conducted to investigate the relationship between teachers' selfefficacy beliefs and job satisfaction levels with meta-analysis method, individual effect sizes and overall effect size of 35 studies were calculated in accordance with the selection criteria. In addition, it was determined whether there was a significant difference between the effect sizes of the studies examining the relationships between teachers' self-efficacy beliefs and job satisfaction levels according to the moderator variables (type of publication, type of school, place of study, type of scales used in the study).
According to the results of the Rosenthal's Safe N Test, Begg and Mazumdar Rank Correlations and Egger Linear Regression method used to determine the validity of the research and publication bias, it was concluded that the publication bias wass low. Before combining the effect sizes of the studies included in the meta-analysis, the normal distribution of the effect sizes was examined, and the homogeneity test was performed to determine which meta-analysis model should be combined with the effect sizes. As a result of the homogeneity test, the meta-analysis model of the study was determined as random effects and the effect sizes of the studies included in the meta-analysis were combined in this model and the overall effect size was calculated.
The first finding of the study was related to the frequency and percentage distributions of the moderator variables of the studies included in the meta-analysis. Studies that examined the relationship between teachers' self-efficacy beliefs and job satisfaction levels were found to be mostly in the type of paper (n = 26, 74,29%). It was determined that more studies were conducted in 2018 (n = 8, 22,86%) than other study years included in the meta-analysis. Number of studies conducted outside Turkey (n = 21, 60%) was observed to be higher than studies conducted in the Turkey (n = 14, 40%). It was determined that most of the studies were performed on high school sample (n = 9, 25,71%). In order to measure the self-efficacy beliefs of the teachers, it was concluded that the researchers used the Teacher Self-Efficacy Scale, which was developed by Tschannen Moran and Woolfolk Hoy (2001), and they used the Minnesota Job Satisfaction Scale, which was developed by Weiss et al. (1967) to measure the job satisfaction levels of teachers.
The second finding of the study showed that there was a moderate and positive relationship between teachers' self-efficacy beliefs and job satisfaction levels. This finding is similar to the metaanalysis findings of Judge and Bono (2001). It is known that the results of the studies not included in the meta-analysis confirm the relationship between teachers' self-efficacy beliefs and job satisfaction levels as they do not meet the selection criteria (Buluç & Demir, 2015;Klassen et al., 2009;Klassen & Chiu, 2010;Redfern, 2016;Türkoğlu et al., 2017;Yıldırım, 2015). Self-efficacy beliefs affect the individual's feelings, thinking style and behaviors (Bandura 1997). Job satisfaction, expressed as an emotional reaction to work, is expected to be closely related to the self-efficacy belief of the individual. However, the fact that the relationship between the two variables is not strong makes us think about what studies can be done to strengthen this relationship. According to Canrinus et al. (2012), self-efficacy, job satisfaction, motivation and commitment are indicators of teachers' professional identity. The way to shape a teacher's professional identity is to influence these indicators. Teachers' clasroom self-efficacy and the satisfaction in the relationship with the team members plays an important role in influencing these indicator. Relationship satisfaction could be strengthened by providing or enhancing a supportive environment, making sure that teachers feel that they are listened by the school board and developing a strong feeling of relatedness between team members. Strengthening these aspects will increase teachers' classroom self-efficacy as well (Asthon & Webb, 1986, as cited in Canrinus et al., 2012. The nature of the relationship between teachers in the school setting determines the professional efforts and performances of teachers, and job satisfaction (Tabancalı, 2016). Supporting social friendship relations and preventing teachers from loneliness will have positive effects on performance and job satisfaction. In a school where social relations are strong, teachers can exchange ideas with their colleagues to help them develop themselves in areas where they are inadequate. In this way, teachers who have the opportunity to develop themselves professionally are likely to increase their self-efficacy beliefs.
The third finding of the study revealed that there was a significant difference between the effect sizes of the studies examining teachers' self-efficacy beliefs and job satisfaction levels according to the moderator variables (type of study, type of school, place of study, type of school and scale used in the study). As any meta-analysis studies conducted in or outside Turkey on the effect of moderator variables on self-efficacy and job satisfaction relation can not be accessed, the results of the study could not be compared with the findings in the literature. However, we can explain the results of individual studies in the literature by comparing with the results of meta-analysis.
Related to the type of publication of the studies; the effect size value of the thesis type (0,411) was found to be greater than the effect size value (0,377) of the paper type. However, this result is not statistically significant. In the meta-analysis studies, the value of p is essentially dependent on two things: the size of the effect and the size of the sample. The calculation of the large effect size for the small sample or the calculation of the small effect size for the large sample may cause the difference between the effect sizes to be significant (Coe, 2002). When the studies included in the meta-analysis are examined in terms of effect size and sample, it can be said that there is a situation in both types of publications, but these studies are not large enough to make a statistically significant difference in effect sizes.
It is important for the thesis to be new in terms of subject, or to re-examine a subject that has been studied previously by using different patterns, sampling methods and measurement tools. In the literature, there are many studies examining the relationship between teachers' self-efficacy beliefs and job satisfaction levels. Therefore, it is expected that the postgraduate theses on this subject will differ from the existing research. It is seen that the measurement tools of the postgraduate theses included in the meta-analysis are examined, especially the scales have variety in terms of job satisfaction. According to İnal (2013), the properties of the tests used in the studies are of great importance. Coe (2002) argues that the reliability of the measurement tool is affected by the effect size and that almost always the instrument containing more items will be more reliable. If two scales whose scores were converted into a percentage, the standard deviation of the percentages of the scale containing more items would be lower than the scale containing the less items. Therefore, although the actual effect is the same, the calculated effect size will be different. Coe (2002) also stated that in the meta-analysis studies based on correlation, correction of the correlation (such as rounding to a desired decimal place) affects the effect size. The differences in the number of items used in the studies and the reliability of the measurement tools used in the studies may have caused significant differences between the effect sizes of the studies according to the type of measurement tool. The same situation may have been effective on the difference between the effect sizes compared to the year of studies. Acoording to the year of studies, the greatest effect size value (0,821) for 2016 and the lowest effect size value (0,299) for 2015 were seen. It is noteworthy that in the successive years, there is a significant difference between the effect sizes of the studies on the subject. A more detailed examination of the characteristics of the measurement instruments of the studies carried out in these years may reveal the reasons behind the emergence of this result.
Related to study location, the effect size of the studies conducted outside Turkey (0,400) was found to be larger than the effect sizes of the studies conducted in Turkey (0,331). To interpret this finding of the research, it may be useful to refer factors that might affect the teachers' self-efficacy beliefs and job satisfaction levels in Turkey. Teachers who start working with specific qualifications may need to update their professionalcompetences or develop new qualifications in accordance with the conditions of the day. There is no system in Turkey to determine how effective teachers' professionalcompetences are to train human profile needed. Teachers' competencies are being developed through in-service trainings. Not all teachers benefit from in-service training as well as the effectiveness of in-service training are arguable. Studies show that there are questions about the prevalence and effectiveness of in-service training being implemented, and that the training is insufficient to achieve targeted change (Demirtaş, 2008;Öztürk, 2003;Yaylan & Sayın, 2006). In-service training programs have some problems in terms of planning, content and evaluation (Demirtaş, 2008). In the in-service trainings, only theoretical knowledge is insufficient to provide new knowledge to teachers (Baykan & Oktay, 2016). The results of a survey conducted by the Ministry of Education in 2006 pointed out that there were problems in the effectiveness of in-service training. In this study, it was determined that 77% of the school administrators and 36,5% of the teachers only participated in the inservice training or not at all. Again from this research, it was determined that subjects considered important by the instructor were not focused while determining their educational needs, and the evaluation of the programs was not done enough; therefore, it was concluded that the feedback that could be used in the development of the programs were not obtained sufficiently (Yayla & Sayın, 2006). As a consequence of this situation, it is possible for teachers to have problems of updating and improving their current professional competences and this is reflected in their self-efficacy beliefs.
There is a consistent relationship between the level of dignity of the profession and the level of the job satisfaction. It is seen that the job satisfaction of the employees who are perceived as more respectable in the social terms is higher (Davis, 1988). In society, there are three important factors that reduce or raise the status of occupations. These are the need for that profession in society, the high standards of living provided to the owner of this profession and the respect for the profession (Gökçe, 1984). Although the status of the teaching profession varies according to societies, it used to have a high dignity in all societies in the past (Tezcan, 1996). But in literature, recent stduies show that there is a decline in the status of the teaching profession in Turkey (Aydın, Demir, & Erdemli, 2015;Gök, 2003;Ozankaya, 2002;Özpolat, 2002;Ünsal & Bağçeci, 2016). The high number of prospective teachers, the lack of criteria for entry to teacher training departments, the lack of successful education of teachers, the lack of undergraduate education, the failure to understand the value of the profession, the failure of the teachers to follow the professional development, the lack of performance, the lack of sufficient care for the profession are evaluated as the reasons that lower the status of the teaching profession (Ünsal, 2018). The fall in the status of the teaching profession based on these reasons may have influenced teachers' job satisfaction levels.
According to the type of school where the study was conducted, the greatest effect size value (0,436) was calculated for primary school and the lowest effect size value (0,169) was calculated for secondary school. The results of the studies on the subject in primary schools show that both the selfefficacy beliefs and the job satisfaction levels of the primary school teachers are higher than the branch teachers (Benzer, 2011;Çimen, 2007;Gençtürk, 2008;Ültanır, 2002;Telef, 2011;Turcan, 2011). The developmental characteristics of the students, the intensity and duration of the relationship with the student may have influenced the teachers' self-efficacy beliefs and job satisfaction levels. In terms of the number of courses and the school year, classroom teachers are more associated with their students than the branch teachers. This association enables more intense communication between the teacher and the student, allowing the teacher to better observe his / her student and to get to know him / her closely. The teacher who knows his students closely can control over the class in every way. The teacher, who controls over his / her class, may feel more sufficient in general (Gençtürk & Memiş, 2010). Classroom teachers, who have the opportunity to observe the development of their students, are more aware of how the students are motivated and how they will be involved in the lesson because they know the personal characteristics of each student more than those of the branch teachers (Çimen, 2007). The fact that classroom teachers have the opportunity to observe the development of their students for five years and to see the developments in them more concrete may have contributed to their higher self-efficacy and job satisfaction (Telef, 2011). Related to the type of self-efficacy scale used in the studies, the greatest effect size value (0,377) was calculated for the scale developed by Tschannen Moran and Woolfolk Hoy (2001); the lowest effect size value (0,206) was calculated for the scale developed by Sherer et al. (1982). During this period, the qualifications that individuals should have differed. Naturally, it is inevitable that the teachers who have an important role in gaining these qualities gained new skills and competences. The development of self-efficacy scales based on relatively more recent competencies may have caused differences in effect sizes.
Related to the type of job satisfaction scale used in the studies, the greatest effect size value according to the type of job satisfaction scale used in the studies (0,518) was calculated for the scale developed by Warr et al. (1979); the lowest effect size value (0,103) was calculated for the scale developed by Spector (1985). Most job satisfaction scales have been developed to assess different aspects of a job. The most commonly used job satisfaction scales measure employee satisfaction through factors such as wages, jobs, audits, promotions, colleagues, awards. Because each job has its own characteristics and it is difficult to measure satisfaction with a general measurement tool. When the most frequently used scales in educational research are analyzed within the scope of meta-analysis, it is seen that these scales are not developed to measure the job satisfaction level of teachers (Giraldo O'Mear, MarinGarcia, & Martinez Gomez, 2014). The development of job satisfaction scales in accordance with the characteristics of the teaching profession may have caused differences in effect sizes.
This study has some limitations. In the studies conducted on the subject, the fact that the relationship coefficient is given on the basis of dimensions rather than the overall scale has limited the number of studies included in the meta-analysis. Since at least two studies are required for each class formed in the calculation of impact size, the 2000, 2006 and 2012 year classes with less than two studies were excluded from the analysis. However, since some years were not conducted in accordance with the selection criteria, the effect sizes of some years could not be calculated. The fact that studies were not conducted on a single school type or that there were less than two studies in some school types led us not to determine whether the effect sizes showed differences according to education levels. The classification of scale types according to the researchers who developed the scale provided information about the countries where the scales were developed. When the countries in which the scales were developed were examined, it was seen that the majority of both self-efficacy and job satisfaction scales were developed in America. This has limited the comparison of research results in a cultural context. Again, the fact that majority of the scales used in studies in Turkey are adapted or that the number of scale developed are not enough for meta-analysis has limited the comparision of scales in a cultural context. Although the study indicates the relationship between self-efficacy and job satisfaction, it does not reveal which self-efficacy dimension is more effective on job satisfaction. Despite these limitations, the study provides new evidence for the relationship between teachers' self-efficacy beliefs and job satisfaction levels. It was found that there was a moderate and positive relationship between teachers' self-efficacy beliefs and job satisfaction levels, and this relationship was significantly different according to study year, study location, type of school and type of scales used in the studies. The meta-analysis of the relationships between the sub-dimensions of self-efficacy and job satisfaction may provide more information on this topic. The effect of the moderator variables such as gender, seniority, branch etc. on relationship between self-efficacy and job satisfaction are not yet known. Future studies or metaanalysis studies that focus on the relevant limitations can make a significant contribution to the literature. The fact that the researchers include statistical information obtained from the overall scale in the studies, the use of the current scale compatible with the cultural context may provide us with more detailed and new information about the relationship between teachers' self-efficacy beliefs and job satisfaction levels.